Gradio Interface  Q&A RAG Chatbot

Combined multiple RAG components  document loaders, text splitters, vector-embedding models, and a vector database into a single pipeline, constructing a fully functional Q&A chatbot under a frontend Gradio Interface. Leveraged LangChain and LLMs to retrieve and answer natural-language questions grounded in content extracted from more than 30 PDF documents.

Natural Language-to-SQL Generator 6 LLM Query Pipeline

Spearheaded backend development, implementing a prompt builder python file that performs basic text processing to create a prompt that writes and returns an SQL query based on real-world home mortgage disclosure act data (HMDA). Engineered a reusable SSH-tunnel and integrated it with the local LLM-Pipeline, enabling several end-to-end submissions of natural-language questions and execution of the appropriate SQL query against a remote Postgres database in the iLab virtual environment.

Multi-Client Interaction  Interactive Chat Server

Spearheaded development of a concurrent, multi-client chat server using TCP sockets and poll()-based I/O multiplexing, supporting 50+ simultaneous connections with minimal CPU overhead. Coordinated with one other team member to develop a lightweight custom protocol for real-time registration, status updates, and private messaging, with structured error handling achieving 100% uptime under multi-client stress testing.

Rising Tides  Regional Risk Analysis

Developed a Java program using a custom 3-layer hierarchical linked list (state  county  community) to structure and analyze regional environmental risk data. Synthesized cross-domain datasets (demographics, pollution, flood risk) to identify correlation factors across hundreds of data points. Designed multi-level traversal and insertion algorithms to reduce search complexity for regional data queries.

✅ Verified by Extern · ⏱️ In progress

Data Analytics, Health Outcomes Externship with ARCeH

Here’s what the data says: put a pin in Bangkok. Draw a five-hour flight radius around it. You just circled half the world's population. And half of those people are kids under five, an age where survival itself isn't guaranteed. Five isn't an arbitrary number: it's the line researchers use as a proxy for a country's overall health, because almost every preventable childhood death, driven by a bad water source, poor air quality, or an infection nobody caught in time, happens before it. Clear five, and the odds of reaching adulthood jump astronomically. Don't, and the cause almost always traces back to poverty, healthcare access, environmental risk, or nutrition, something that could have been mapped, measured, and prevented in time. That's the gap this externship exists to close: turning scattered, messy public data into a specific, defensible answer about what's actually driving that risk in a given place, and pass it to the people at ARCeH who want to know about it.